Professional Documents
Culture Documents
In"nite and periodic arrays of porous disks are considered, in oscillatory #ow perpendicular to
their planes. This con"guration is of interest for the o!shore oil industry, for instance to be used
as dampers to reduce the heave motion of truss SPAR platforms. The hydrodynamic problem is
solved by the method of matched eigenfunction expansions, under the assumptions that losses
of head, proportional to the square of the traversing velocities, occur at the disks, and that
potential #ow theory is applicable. Added mass and damping coe$cients are derived, as
functions of the relative spacings between the disks, and of the parameter (A/a) (1!q)/(2kq),
where A is the #ow motion amplitude, a the disks radii, q the porosity, or open-area ratio, and
k a discharge coe$cient, close to 0.5. Results are also given for disks that are partially porous,
from their axis to some radial distance. The practical applicability of the obtained results is
discussed. 2001 Academic Press
1. INTRODUCTION
OFFSHORE FLOATING OR COMPLIANT STRUCTURES are susceptible to resonant behaviour, under
linear or nonlinear wave excitation. An e$cient means to reduce the amplitudes of the
resonant responses is to increase energy dissipation, by enhancing drag forces on some parts
of the structures. A well-known example is the resonant roll motion of ships and barges,
which bilge keels reduce e$ciently. Another means is to shift the resonant frequencies
outside the wave frequency range, for instance by increasing the added mass through
additional plates.
In the past years some interest has arisen for perforated elements to be used as passive
dampers [e.g. see Downie et al. 2000]. In 1990, the author (Molin & Legras 1990) proposed
a hydrodynamic analysis of the Roseau tower stabilizer, an open box with slots. The
theoretical model was based on potential #ow theory and on the assumption that losses of
head, proportional to the square of the (relative) traversing #uid velocities, are induced at
the openings. Added mass and damping coe$cients were obtained and good agreement was
observed with experimental results from dedicated model tests [see also Damy & Molin
(1991)]. A remarkable feature is that the added mass and damping are amplitude dependent.
Other applications were subsequently considered, such as the heave motion of a semi-
submersible platform with perforated pontoons (Molin 1992a, b), or the damping of waves
through a series of vertical screens (Molin & Fourest 1992), or over a perforated horizontal
plate (Molin & BeH tous 1993; Molin 1999), within all cases there is good agreement between
theoretical results and experiments. The case of steady current #ow over shrouded cylinders
was also investigated (Molin 1993).
An o!shore production system that has gained some success, recently, is the SPAR
platform, consisting in a #oating vertical cylinder, of large draft [e.g. see Le Blanc (1996)].
The heave natural period is roughly equal to 2n(d/g, d being the draft and g the
acceleration due to gravity. As a result, drafts over 200 m are necessary to ensure that the
heave natural period be well beyond the wave period range. At smaller drafts, means to
enhance damping are necessary. A proposed system consists in piling up a series of
horizontal plates below the main hull, linked by a truss structure. The damping e$ciency of
such solid plates has been studied experimentally (Prislin et al., 1998; Lake et al., 1999).
They also provide valuable supplementary added mass.
Another o!shore structure is the TPG 3300 that consists in a deeply immersed #at
pontoon, of large dimensions, supporting three or four vertical columns running through
the free surface up to the deck (Thomas & Gaubil 1997). The pontoon is open in its centre,
to permit passage for the risers. Being able to adjust the heave damping is an attractive issue
that could make viable designs with reduced heave natural periods, as compared to the
present ones. A possibility considered is to partially obstruct the opening through the
pontoon, to reach an optimized open-area ratio.
These two cases have motivated the present study, where we consider an in"nite and
periodic series of disks. The disks are either completely or partially porous. The periodicity
assumption renders the mathematical problem easier to tackle, and gives valuable results
for several plates in proximity. For an isolated disk it su$ces to increase the periodicity
length beyond some distance. The circular disk assumption is also made for convenience:
the geometry becomes axisymmetric.
The theoretical frame is identical with the one used in the quoted previous publications:
use is made of potential #ow theory. In the particular geometry considered here, this
assumption is somewhat criticizable, since the #ow separates not only through the openings
but also at the edges of the disks. This latter e!ect is not taken into account in our
theoretical frame. In the discussion part of the paper, we hint that it becomes negligible at
su$ciently low values of the Keulegan}Carpenter number, expressed as KC"nA/a,
A being the motion amplitude and a the disk radius. It is for these small
Keulegan}Carpenter numbers that porous disks appear as more e$cient dampers than
solid ones.
2. THEORY
2.1. EIGENFUNCTION EXPANSIONS
We consider an in"nite and periodic array of porous disks, with Oz as their axisymmetry
axis; their radius is a, their spacing is l. The array is "xed. The #uid domain is unbounded.
The #ow velocity, far away from the array, is = (t) in the Oz direction.
We use a cylindrical coordinate system (R, h, z) with the plane z"0 conciding with the
porous disk number 0. We will speci"cally consider the cell number 1, in-between the
porous disks 0 and 1 (extending from z"0 up to l in z, and from 0 to a in R). Within this cell
the velocity potential of the #ow will be written as the sum of two eigenfunction expansions.
We also use an eigenfunction expansion in the outer domain R5a.
In this outer domain the #ow velocity is z-periodic with periodicity length l. The velocity
potential is also antisymmetric in z with respect to each plane z"jl. Hence, it can be
decomposed as
K (j R)
u (R, z, t)"= (t)z# A (t) sin j z L , (1)
C L L K (j a)
L L
where j "2nn/l and K is the modi"ed Bessel function of the second kind and order 0.
L
With u thus written, the Laplace equation and the condition at radial in"nity are ful"lled.
C
In the inner domain, two sets of eigenfunctions are used. The "rst one corresponds to the
PERIODIC ARRAYS OF POROUS DISKS 277
case when the disks are solid and the velocity potential satis"es the homogeneous Neumann
condition. The expansion for u is then
G&
I (k R)
u (R, z, t)"B (t)# B (t) cos k z K , (2)
G& H K K I (k a)
K K
where k "(2m!1)n/l and I is the modi"ed Bessel function of the "rst kind and order 0.
K
Only odd values of k l/n are retained because of the antisymmetry property. Note that
K
B takes a di!erent value in each cell. As a matter of fact one obtains readily that
H
B (t)"= (t)(2j!1)l/2, where j is the cell number.
H
The second expansion is given in the following section, after the boundary condition at
the porous walls has been described.
Such comparisons are reported in the quoted papers. They deal with various geometries
and #ow conditions, with porosity ratios in the range 10}40%. Quite favourable agree-
ments between experimental and numerical results are reported, with the k coe$cient
usually slightly above 0)5.
The second condition to impose at the disk results from mass conservation which
requires that
u\"u>"u .
X X X
At the edge of the disk u\"u> and hence the traversing velocity is nil. We take
advantage of this to expand the traversing velocity, u (R, 0, t), as
X
u (R, 0, t)" C (t)J (l R), (7)
X G G
G
where J is the Bessel function of order 0 and the &wavenumbers' l are the roots of
G
J (l a)"0. It can readily be veri"ed (see Appendix A) that the functions J (l R) form an
G G
orthogonal set over the interval [0 a]:
?
J (lGR)J (lHR)R dR,0
for iOj. It can be checked that the set is complete.
When u is thus given at z"0 and l, a particular solution within the cell is
X
cosh l z!cosh l (z!l)
u (R, z, t)" C (t) G G J (l R). (8)
G. G l sinh l l G
G G G
The general solution within the cell is obtained by adding up u and u :
G& G.
I (k R) cosh l z!cosh l (z!l)
u (R, z, t)"B (t)# B (t) cos k z K # C (t) G G J (l R).
G H K K I (k a) G l sinh l l G
K K G G G
(9)
All that remains to be done is to match u and u , and their radial derivatives, for R"a,
G C
and use the discharge equation (6) as an evolution equation to advance the A , B and
K L
C coe$cients in time, given the outer-#ow velocity = (t).
G
and F( j) is the averaged value of u "u " over the interval [R R ]. The radius a has been
X X H H>
discretized in N segments [R R ] to permit a numerical evaluation of the integral
NR H H>
?
J (l R) C (t)J (l R) C (t)J (l R) R dR.
G
G
G G
G
G G
(24)
Q (t)l ?. 2 + ?. I (k R)J (l R)
=
" J (l R)R dR# BQ (t) K I R dR
c I c K I (k a)
I I K K
1 1!q ?.
!
c 2kq
I
J (l R)"u "u R dR
I X X
(30)
3. RESULTS
3.1. SOLID DISKS
First we consider the case of fully solid disks. In such case the problem reduces to equation
(17) where C,0. Once this linear system is solved, with = (t)"1, the added mass M is
?
readily derived as
+ I (k a)
M "on la#4a B K . (32)
? K k I (k a)
K K K
Figure 1 gives the added mass coe$cient C "M /( oa) as a function of the ratio a/l.
? ?
When the relative spacing goes to in"nity (a/lP0), the case of one single disk C " is
?
recovered. When the relative spacing goes to zero, the added mass should reduce to the
mass of the entrapped water: M "onal giving C Knl/a. This asymptotic behaviour is
? ?
well recovered by our numerical results.
For these calculations the truncation orders M and N of the expansions 1 and 2 were
taken equal to 100.
Now we investigate the rate of convergence with respect to the truncation order. This is
shown in Figure 2, which gives the added mass coe$cient for l/a"2 and 10 versus 1/M (N
being taken equal to M). It can be seen that convergence is quicker for l/a"2.
Figure 1. Solid disk. **, Added mass coe$cient as a function of a/l; } } }, nl/a.
282 B. MOLIN
Figure 2. Solid disk. Added mass coe$cient as a function of 1/M : **, l/a"2; } } }, l/a"10.
Figure 3. Porous disk in sinusoidal #ow. **, Simulated hydrodynamic force F(t) (normalized by
oaAu) versus nondimensional time t/¹; } } } }, sin (ut); ----, cos (ut).
Figure 4. Porous disk in sinusoidal #ow. Left- and right-hand sides (normalized by oaAu) of the
discharge equation (6) at two di!erent instants: **, left-hand side; ----, right-hand side.
We now give the added mass and damping coe$cients, when the parameter
(A/a)(1!q)/(2kq) varies from zero up to ten. They are given in Figures 5}7 for l/a"1, 2
and 5 (results for l/a'5 di!er little from this latter case).
When the parameter (A/a)(1!q)/(2kq) goes to zero both the added mass and damping
coe$cients become nil: the water particles #ow freely through the porous wall. When it
increases to in"nity, the solid disk results are asymptotically recovered, with the damping
coe$cient going to zero and the added mass one increasing up to its value given in Figure 1.
For a value of (A/a)(1!q)/(2kq) somewhere in-between 1 and 1)5 (for the relative spacings
considered here) the damping coe$cient reaches its peak value. It can be observed that it is
284 B. MOLIN
Figure 5. Porous disk in sinusoidal #ow. Relative spacing l/a"1. Added mass and damping
coe$cients versus (A/a)(1!q)/(2kq): **, added mass coe$cient C ; -----, damping coe$cient C .
? @
Figure 6. Porous disk in sinusoidal #ow. Relative spacing l/a"2. Added mass and damping
coe$cients versus (A/a)(1!q)/(2kq): **, added mass coe$cient C ; } } }, damping coe$cent C .
? @
Figure 7. Porous disk in sinusoidal #ow. Relative spacing l/a"5. Added mass coe$cient (**) and
damping coe$cient (} } }) versus (A/a)(1!q)/(2kq).
PERIODIC ARRAYS OF POROUS DISKS 285
then equal to the added mass coe$cient, that is the hydrodynamic force is 453 out of phase
with the #ow acceleration. Similar features are obtained with porous circular cylinders
(Molin 1992b). Another remarkable result, also obtained with porous cylinders and with
2-D plates (Molin 2000), is that the peak value of the damping coe$cient is always equal, or
very close, to half the value of the added mass coe$cient in the limit (A/a)(1!q)/
(2kq)PR, that is for the solid case.
Coming now to the e!ect of relative spacing, it can be seen from Figures 5}7 that, when
l/a varies from 1 to 5, the peak value of the damping coe$cient increases by less than 40%.
To maximize the damping e!ect over a given length, it is thus preferable to install many
plates.
Figure 8. Partly porous partly solid disk in sinusoidal #ow. Traversing velocity </(Au) along the
radius at di!erent instants. Relative spacing l/a"5. Radius ratio a /a"0)5.
.
286 B. MOLIN
Figure 9. Partly porous partly solid disk in sinusoidal #ow. Pressure di!erential Dp/(oaAu) along
the radius at di!erent instants. Relative spacing l/a"5. Radius ratio a /a"0)5.
.
Figure 10. Partly porous partly solid disk in sinusoidal #ow. Added mass (C ) and damping
?
(C ) coe$cients versus (A/a)(1!q)/(2kq). Relative spacing l/a"5. Radius ratio a /a"0)5: **,
@ .
C ; } } }, C .
? @
the area), the ratio is 85%. This is because blockage e!ects strongly enhance the #ow
velocity through the porous part.
4. DISCUSSION
Our theoretical approach, based on perfect #uid and irrotational #ow assumptions, predicts
no damping for a solid disk. Practically however, the #ow separates at the edge, inducing
drag forces and energy dissipation. With a porous disk, there is separation at the edge as
PERIODIC ARRAYS OF POROUS DISKS 287
Figure 11. Partly porous partly solid disk in sinusoidal #ow. Added mass (**) and damping (} } })
coe$cients versus (A/a)(1!q)/(2kq). Relative spacing l/a"5. Radius ratio a /a"0)75.
.
well, and also separation through the openings. Our theoretical model takes the latter e!ect
into account (in a somewhat idealized way), but not the former one. One feels intuitively
that our model may yield appropriate values for the damping in so far as the e!ect of
separation through the openings is dominant with regard to the e!ect of separation at the
edge. A way to check whether this can be the case is to compare the damping values that we
obtain for a porous disk with available experimental values of the damping for a solid disk.
If the porous disk damping exceeds the solid disk damping by far, then there is a chance that
our model could be of some value.
Experimental data for solid disks are provided by Lake et al. (1999). They obtain that the
damping coe$cient varies linearly with the Keulegan}Carpenter number KC, for
0)14KC40)5. From their "gure 5, one has, roughly :
3B A
B"0)25KC, where B" , KC"n .
16oau a
Referring to our Figure 7, we see that with an optimized porosity ratio, the damping
coe$cient B/(oau) is equal to about 1)25. Hence, the ratio (optimized porous damp-
ing)/(solid damping) is given by
B
K0)95KC\.
B
When the KC number is larger than one, no extra damping can be gained by making the
disk porous. There is little point in wondering whether our theory is applicable or not. One
has to better keep the disks solid.
The matter becomes di!erent if the KC number is much smaller than one. For SPAR
platforms it has been considered to add up solid plates to the truss structure below the hull,
in order to increase the damping in heave. Typically, the SPAR radius is around 15 m and
the motion amplitude is small, of the order of 1 m. This means a KC number around 0)2,
assuming the plates radii to be equal to the SPAR radius. Then there is a lot to gain by
making the plates porous. With a porosity slightly below 20%, the heave damping can be
increased by a factor 4 or 5. This result would demand con"rmation through dedicated
model testing.
288 B. MOLIN
In Molin & Legras (1990) a somewhat similar situation is encountered. The structure
considered is a truncated, open-ended, vertical cylinder, initially designed to increase the
horizontal added mass of compliant towers. Forced oscillation tests are reported on
di!erent models, of height and diameter both equal to 1 m, at KC numbers ranging from
nearly zero up to about 0)5. Porosity ratios of 10, 20 and 24% were achieved, either as
horizontal slots (12 over the height of the cylinder) or with small circular perforations. It
was found that the measured and calculated damping coe$cients agreed closely at the
smallest porosity ratio, for KC numbers ranging from zero up to the value associated with
the peak in the damping curve. At larger KC numbers they started to deviate from each
other, the experimental values remaining more or less constant while the numerical ones
decreased slowly. This discrepancy was attributed to the drag forces taking place at the
lower and upper sharp edges of the truncated cylinder.
Another remarkable result obtained here is the sensitivity of the added mass coe$cient of
a porous disk to the porosity ratio and Keulegan}Carpenter number. It may also look
surprising that the added mass becomes nil when the KC number goes to zero. For the
perforated truncated cylinder considered in Molin & Legras (1990) this behaviour was
con"rmed by the experimental results. In that case, the radius of the circular perforations
was equal to 1 mm, to be compared with a cylinder diameter of 1 m and motion amplitudes
being as small as 2 mm. In the slotted version, owing to the larger spacings in-between the
slots, the added mass coe$cient did not quite decrease to zero as the motion amplitude was
reduced, but it nevertheless became very small, amounting to only about 10% of the zero
porosity case.
The sensitivity of the added mass coe$cient with the motion amplitude o!ers interesting
perspectives when one is concerned with resonance problems. They are discussed in Molin
(1992b) in the particular case of semi-submersible platforms with slotted pontoons.
ACKNOWLEDGEMENTS
The incentive to work on this problem was provided through discussions with Prof. J.M.R.
Graham of Imperial College, and with Ph. Weber of Technip.
REFERENCES
DAMY, G. & MOLIN, B. 1991 Stabilizers for compliant towers: numerical and experimental studies. In
Proceedings Brasil Owshore+91, Rio de Janeiro.
DOWNIE, M. J., GRAHAM, J. M. R. & WANG, J. 2000 The e!ectiveness of porous damping devices. In
Proceedings 10th International Owshore and Polar Engineering Conference (ISOPE), Vol. III, pp.
418}425, Seattle.
LAKE, M., HE, H., TROESCH, A. W., PERLIN, M. & THIAGARAJAN, K. P. 1999 Hydrodynamic coe$cient
estimation for TLP and SPAR structures. In Proceedings 18th International Conference
on Owshore Mechanics and Arctic Engineering (OMAE), paper 4103, St. John's. Newfoundland,
Canada.
LE BLANC, L. 1996 Spar-shaped drilling unit designed for 8,000}10,000 ft depth corridor. Owshore 56,
30}31.
MOLIN, B. 1992a A new small semi concept for marginal "elds. In Proceedings 24th Owshore
Technology Conference, paper 7050, Houston, Texas, U.S.A.
MOLIN, B. 1992b Motion damping by slotted structures. In Hydrodynamics: Computations, Model Tests
and Reality, Developments in Marine Technology, Vol. 10, pp. 297}303. Dordrecht: Elsevier.
MOLIN, B. 1993 A potential #ow model for the drag of shrouded cylinders. Journal of Fluids
& Structures 7, 29}38.
MOLIN, B. 1999 Numerical and physical tanks. Making them "t, 22nd Memorial Georg Weinblum
Conference, Hamburg (to appear in Ship Technology Research).
MOLIN, B. 2000 Added mass and damping of two-dimensional porous plates, Unpublished note.
PERIODIC ARRAYS OF POROUS DISKS 289
MOLIN, B. & LEGRAS, J.-L. 1990 Hydrodynamic modelling of the Roseau tower stabilizer. In Proceed-
ings 9th International Conference on Owshore Mechanics and Arctic Engineering (OMAE), Houston,
¹X, ;.S.A. Vol I, Part B, pp. 329}336.
MOLIN, B. & FOUREST, J.-M. 1992 Numerical modeling of progressive wave absorbers. In Proceedings
7th International Workshop on Water Waves & Floating Bodies, pp. 199}203, Val de Reuil, France.
MOLIN, B. & BED TOUS, P. H. 1993 AtteH nuation de la houle par une dalle horizontale immergeH e et
perforeH e. In Actes des QuatrieH mes JourneH es de l'Hydrodynamique, pp. 387}400, Nantes, France (in
French).
PRISLIN, I., BLEVINS, R. D. & HALKYARD, J. E. 1998 Viscous damping and added mass of solid square
plates. In Proceedings 17th International Conference on Owshore Engineering and Arctic Engineer-
ing (OMAE), Lisbon, Portugal.
THOMAS, P. A. & GAUBIL, F. 1997 From the TPG 500, production drilling quarter jack-up to the TPG
3300, semi-submersible platform for deep water. In Proceedings 29th Owshore Technology Confer-
ence, Houston, ¹X, ;.S.A., paper 8564.
APPENDIX A: INTEGRALS
A.1. ORTHOGONALITY OF THE J (lGR) FUNCTIONS OVER [0 a]
Let u(R)"J (lGR) and v(R)"J (lHR) with lGOlH; u and v satisfy the di!erential equations
Ru#u#lG Ru"0,
Rv#v#lRv"0.
H
Multiplying both sides of the "rst equation with v, both sides of the second one with u,
and subtracting them, one gets
*
[R(uv!uv)]#(lG !lH )Ruv"0,
*R
Hence
? ?
(lG !lH ) uvR dR"!R(uv!uv) "0
if u(a)"v(a)"0.
J (lGR)R dR
?
A.2. INTEGRAL
* z
(J (z)#J (z)) "zJ (z).
*z 2
Since J (lGa),0, this leads to
? a
J (l R)RdR" J (l a)
G 2 G
I (kKR)J (lGR)R dR
?
A.3. INTEGRAL
Rv#v!kKRv"0.
Multiplying the "rst equation with v, the second one by u, subtracting, and integrating,
gives
? 1 ?
Ruv dR" R(uv!uv) .
lG #kK
One obtains "nally